Prioritizing public investments for low-carbon rural e-commerce in China: an AHP–TOPSIS framework linking income growth and environmental performance

Abstract

Introduction:

Rural e commerce is widely promoted in China to expand market access and increase rural incomes. However, its rapid development also creates environmental pressures through logistics related emissions and packaging waste. This study aims to identify and prioritize public investment packages that can support rural income growth while promoting low carbon development.

Methods:

An integrated Analytic Hierarchy Process and Technique for Order Preference by Similarity to Ideal Solution (AHP TOPSIS) framework is developed to evaluate and rank public investment options for rural e commerce development. Pairwise comparison questionnaires from 60 experts are aggregated using the geometric mean to derive criterion weights. Six main criteria and 20 indicators are considered, covering digital infrastructure, logistics capability, human capital, market linkage, inclusive economic impact, and environmental outcomes. Five alternative public investment packages are evaluated using TOPSIS based on their distances from the positive and negative ideal solutions. Sensitivity analysis with ±30% weight perturbations and scenario analysis under inclusion, growth, and environmental priorities are also conducted to assess ranking robustness.

Results:

The baseline TOPSIS results show that A4 ranks first with a closeness coefficient of 0.630, followed by A2 with 0.593. A5 and A1 form a middle tier with coefficients of 0.556 and 0.534, respectively, whereas A3 ranks last with 0.489. The sensitivity analysis shows limited rank reversals, mainly among the middle ranked alternatives. Across the inclusion, growth, and environmental scenarios, A4 and A2 consistently remain the two highest ranked alternatives.

Discussion:

The results indicate that skills and entrepreneurship support (A4) and low carbon logistics investment (A2) provide the most robust performance across different policy priorities. These findings suggest that rural development policies should combine human capital development with low carbon logistics infrastructure rather than focus on income growth alone. The proposed AHP-TOPSIS framework provides a transparent and robust decision support tool for prioritizing rural e commerce investments under multiple economic, social, logistical, and environmental objectives.

1 Introduction

Rural ecommerce has become a central pathway through which China seeks to broaden market access for rural producers, raise household incomes, and support rural revitalization (Zhang et al., 2024; Zhang et al., 2022; Liu and Zhou, 2023). Empirical evidence increasingly shows that rural ecommerce participation and related policy programs can generate measurable income benefits and distributional improvements (Luo et al., 2023; Chen and Pan, 2019). For example, research using Taobao village data links e-commerce participation to household income growth and local spillovers through service networks and scale effects (Liu and Zhou, 2023; Luo and Niu, 2019; Qiao, 2025). Recent policy evaluation studies also report that rural e-commerce programs can increase farmers’ incomes while reducing income inequality, using quasi-experimental designs and panel household datasets (Zhang Y. et al., 2025; Lv et al., 2025).

At the same time, the rapid growth of e-commerce puts pressure on the environment through transport activity, packaging materials, and waste disposal (Zhang M. et al., 2025; Ding et al., 2025). Life cycle-oriented studies estimate substantial carbon footprints associated with e-commerce express packaging and show strong spatial concentration and embodied transfer patterns across provinces (Luo and Zhu, 2025; Park et al., 2010). Related evidence suggests that rural e-commerce expansion can increase pollution drivers through transport and rural waste growth, which highlights the need to integrate economic and environmental objectives in investment planning (Tang et al., 2025; Yu et al., 2024).

From a policy and management perspective, local governments face a practical allocation problem (Hu et al., 2026; Hu et al., 2025). Budget constraints require prioritizing among competing investment packages, such as digital connectivity upgrades, last-mile logistics improvements, cold chain infrastructure, skills and entrepreneurship support, market linkages, and standardization measures (Jiang and Yuan, 2026; Lu and Yoon, 2026). These interventions operate through different mechanisms and produce trade-offs (Sun and Abd Kadir, 2026; Jiang et al., 2026). Connectivity and skills may unlock participation and productivity gains, yet logistics and the cold chain shape delivery efficiency, product loss, and market reach, while standardization and traceability affect access to higher-value markets and compliance (Tang et al., 2025; Bachofner et al., 2022; Nwabekee et al., 2023).

Meanwhile, low-carbon logistics and packaging waste management can reduce emissions intensity and waste burdens but may have different short-run costs and implementation constraints (Tang et al., 2018; Chu et al., 2021). The decision challenge is therefore multidimensional and calls for a transparent method that can combine expert knowledge about relative importance with observed performance information across locations.

1.1 Literature review

Existing studies have examined rural e-commerce from both economic and environmental perspectives, while multi-criteria decision-making methods such as AHP and TOPSIS have been widely applied to evaluate complex policy alternatives across multiple objectives (Zhang Y. et al., 2025; Li and Qin, 2022; Kang et al., 2021; Kumar et al., 2020; Kumar et al., 2022; Kumar et al., 2023; Kumar et al., 2025). However, their combined application for prioritizing low-carbon rural e-commerce investments in China remains limited. Empirical studies have reported positive effects of rural e-commerce on household income and rural welfare, although the magnitude of these effects differs across regions and stages of development (Zhang Y. et al., 2025; Li and Qin, 2022; Qiuyun et al., 2025). Evidence from Taobao Village development links e-commerce clustering to rural income gains, with stronger effects in specific geographic contexts, while policy evaluation work using household panel data finds that national rural e-commerce initiatives can improve welfare outcomes, such as reducing rural income inequality (Zhang Y. et al., 2025; Li and Qin, 2022).

At the same time, recent findings highlight trade-offs in labor allocation and income composition, suggesting that complementary investments in capabilities and market access can determine whether e-commerce expansion translates into broad-based gains. These results support an evaluation design that distinguishes enabling conditions from final outcomes and allows priorities to differ across localities (Qiuyun et al., 2025).

Recent life cycle assessment studies have quantified greenhouse gas emissions and packaging waste associated with e-commerce logistics and express delivery systems in China (Su et al., 2020; Zhou et al., 2024; Kang et al., 2021). Life cycle assessments of express packaging in China identify substantial contributions to greenhouse gas emissions across upstream material production and downstream disposal stages, and more recent work continues to document the scale of packaging waste and the importance of recycling and material substitution pathways (Su et al., 2020; Zhou et al., 2024). Sector-level analyses of express delivery emissions further indicate that decarbonization requires both operational efficiency improvements and the adoption of low-carbon technologies, reinforcing the need to include indicators of delivery emissions intensity and packaging pressure when evaluating public investment options (Kang et al., 2021).

Recent empirical studies have reported positive relationships among rural e-commerce, rural income growth, industrial upgrading, and low-carbon development, although the magnitude and persistence of these effects vary across studies (Ji et al., 2023; Du et al., 2023; He et al., 2025). Ji et al. (2023) reported that rural e-commerce reduced agricultural carbon emissions by approximately 14.4% through agricultural economic growth, industrial upgrading, and productivity improvements (Ji et al., 2023). However, the authors also found that the carbon-reduction effect was not long-lasting and varied significantly across regions, with stronger impacts observed in eastern China and in non-major grain-producing areas. Similarly, Du et al. (2023) found that rural e-commerce demonstration policies reduced carbon emissions through improved transportation efficiency and digitalization (Du et al., 2023). More recent evidence by He et al. (2025) suggested that e-commerce development promotes low-carbon development through green innovation, industrial restructuring, and resource-sharing mechanisms (He et al., 2025). Nevertheless, the literature does not fully agree on whether these benefits are universal because several studies emphasize that outcomes depend heavily on local infrastructure conditions, digital readiness, and market maturity.

Previous studies report mixed evidence regarding the environmental effects of rural e-commerce expansion. Some studies reported that digital platforms improve logistics efficiency and reduce carbon emissions (Du et al., 2023; He et al., 2025; Hu and Yin, 2026), whereas others highlighted increases in packaging waste, transportation demand, and energy consumption associated with rapid e-commerce growth (Su et al., 2020; Zhou et al., 2024; Kang et al., 2021). On the one hand, recent studies argue that digital platforms can improve logistics efficiency, optimize resource allocation, and contribute to carbon reduction. For example, Hu and Yin, (2026) found that rural e-commerce significantly reduced carbon emissions by narrowing the urban-rural income gap, promoting industrial upgrading, and improving digital infrastructure (Hu and Yin, 2026). On the other hand, several researchers highlight that rapid e-commerce growth can increase environmental pressures through packaging waste, transportation demand, energy-intensive digital infrastructure, and increased consumption. Evidence from recent studies indicates that carbon emissions may initially increase during the early stages of e-commerce expansion before declining as logistics systems and digital facilities become more efficient. These mixed findings reveal that the relationship between rural e-commerce and environmental sustainability remains complex and context-dependent. Furthermore, existing studies primarily focus on estimating economic or environmental impacts, while limited research provides a practical decision-support framework for prioritizing alternative public investment packages that simultaneously address economic, social, and environmental objectives. This gap motivates the development of the integrated AHP–TOPSIS framework proposed in the present study.

Previous studies have applied multi-criteria decision-making methods to evaluate policy and infrastructure alternatives involving economic, environmental, and social objectives (Kumar et al., 2020; Kumar et al., 2022; Kumar et al., 2023; Kumar et al., 2025). Integrated AHP-TOPSIS models are widely used to structure expert judgments and rank alternatives in logistics and sustainability-related decisions, including supplier evaluation and infrastructure prioritization, where criteria span economic and environmental dimensions. However, few studies have integrated economic and environmental evidence on rural e-commerce into a unified AHP-TOPSIS framework to prioritize alternative public investment packages in China. This motivates the present framework, which connects the empirical mechanisms identified in the literature to an operational weighting and ranking procedure that can be compared across case areas and tested for robustness under alternative policy priorities.

Transportation efficiency corresponds to the logistics capability dimension, which receives a relatively high weight and helps explain why A2 performs strongly in the baseline and ranks first in environment-oriented scenarios. The stronger effects in more digitally advanced areas also support the interpretation that A1 digital infrastructure upgrades can be a high priority in less connected regions, even if they are not top-ranked in the overall baseline, because digital readiness conditions the ability of logistics and market linkage measures to translate into emissions reductions.

Although China provides an important empirical context due to its large-scale rural e-commerce expansion and extensive rural revitalization initiatives, similar challenges have been reported in many other developing economies. Studies from countries such as India, Indonesia, Vietnam, Kenya, and Bangladesh indicate that rural producers frequently face constraints related to digital connectivity, logistics infrastructure, market access, entrepreneurial skills, and environmental sustainability. While institutional settings differ across countries, the fundamental policy challenge of prioritizing limited public resources among competing investments remains largely similar. Consequently, the evaluation dimensions considered in this study, including digital infrastructure, logistics capability, human capital, market linkage, inclusive economic impact, and environmental outcomes, are not unique to China. The proposed AHP–TOPSIS framework, therefore, possesses broader applicability and can be adapted to support investment prioritization and rural digital development strategies in other developing-country contexts, subject to local data availability and policy objectives.

1.2 Objectives and novelty

Existing studies have extensively examined the economic benefits of rural e-commerce and, more recently, its environmental implications. However, the literature remains fragmented between studies evaluating economic outcomes and those assessing environmental impacts. Furthermore, most existing research focuses on measuring the effects of rural e-commerce policies rather than on providing practical tools for prioritizing public investment options across multiple, potentially conflicting objectives. As a result, limited attention has been paid to identifying which investment package to prioritize when economic growth, social inclusion, and environmental sustainability must be considered simultaneously.

To address this gap, this study aims to develop and apply an integrated AHP–TOPSIS framework for prioritizing low-carbon rural e-commerce investment alternatives in China.

Based on these objectives, the study develops an integrated AHP–TOPSIS framework to support the evidence-based allocation of rural development resources and to help policymakers identify investment priorities that simultaneously promote income growth and low-carbon development.

2 Methodological frameworks

A structured multi-criteria decision-making framework is adopted to integrate AHP and TOPSIS to derive an overall priority ranking of alternatives under multiple, potentially conflicting criteria. The workflow, summarized in Figure 1, proceeds through a sequence of logical steps designed to ensure transparency, internal consistency, and replicability. First, the decision problem is specified by defining the evaluation objective and the set of alternatives to be assessed. Second, an evaluation index system is constructed by identifying criteria and subcriteria that capture the relevant dimensions of performance.

FIGURE 1

For each indicator, the measurement method and preference direction are clarified, and any necessary preprocessing is performed to ensure comparability, including unit harmonization and orientation adjustment so that all indicators follow a consistent benefit interpretation. Third, AHP is applied to determine the criteria weights based on expert elicitation. Pairwise comparison matrices are formed for each level of the hierarchy, and normalized priority vectors are obtained to represent relative importance. Consistency is then evaluated using standard AHP diagnostics to verify that the judgments are coherent and acceptable for weight extraction. The resulting weight set is interpreted as the decision makers’ preference structure and is carried forward to the ranking stage. Fourth, TOPSIS is used to synthesize weighted performance across indicators and generate the final ordering of alternatives.

A decision matrix is assembled from the indicator data, normalized to remove scale effects, and multiplied by the AHP-derived weights to obtain a weighted normalized matrix. Positive ideal and negative ideal solutions are then identified to represent the best and worst attainable performance across criteria. Distances from each alternative to these ideal points are computed, and a closeness coefficient is calculated to quantify how near each alternative is to the positive ideal while remaining far from the negative ideal. Alternatives are ranked in descending order of the closeness coefficient, providing the primary decision support output of the framework.

Finally, the reliability of the ranking can be examined by testing the stability of results under plausible variations in weights and input values, allowing the decision maker to identify criteria that most influence the final ordering and to assess the robustness of the recommended priorities. Figure 1, therefore, serves as the methodological roadmap linking problem definition, weight derivation, and TOPSIS-based ranking into a single integrated AHP TOPSIS workflow.

The methodological framework adopted in this study is grounded in recent advances in multi-criteria decision-making and sustainable supply chain performance evaluation. Kumar et al. (2020) developed a comprehensive framework for assessing cold supply chain performance in vaccine distribution and demonstrated the importance of integrating multiple economic, operational, and sustainability criteria into a structured evaluation system (Kumar et al., 2020). Building on this perspective, Kumar et al. (2022) further highlighted the role of decision-support frameworks in identifying and prioritizing improvement strategies under complex supply chain environments (Kumar et al., 2022). Similarly, Singh et al. (2026) emphasized the need for technology-oriented assessment approaches to evaluate sustainability and environmental performance, particularly amid growing concerns about global warming and carbon emissions. More recently, Kumar et al. (2023) proposed an interval-valued Pythagorean fuzzy decision-support framework for enhancing food quality through cold supply chain performance assessment, illustrating the growing application of advanced MCDM techniques in sustainability-oriented decision problems (Kumar et al., 2023; Kumar et al., 2025). Although these studies focus on cold supply chain systems, their methodological foundations are directly relevant to the present research, as both contexts involve evaluating alternative strategies across multiple, potentially conflicting economic, operational, and environmental criteria. Therefore, the integrated AHP–TOPSIS framework employed in this study extends these established methodological principles to prioritize low-carbon rural e-commerce investment alternatives in China.

Although several multi-criteria decision-making (MCDM) methods, such as ANP, VIKOR, ELECTRE, and PROMETHEE, have been widely applied in sustainability and infrastructure planning, the integrated AHP–TOPSIS approach was selected because it is particularly suitable for the objectives and data structure of this study. AHP provides a systematic mechanism for deriving criterion weights from expert judgments while incorporating consistency testing to ensure the reliability of pairwise comparisons.

However, AHP alone is primarily designed for weight determination and becomes less efficient when ranking multiple alternatives across many indicators, as it requires extensive pairwise comparisons among alternatives. Therefore, TOPSIS was integrated with AHP to overcome this limitation. TOPSIS enables the efficient ranking of alternatives based on their relative distances from positive and negative ideal solutions and can simultaneously handle numerous quantitative and qualitative indicators. Compared with more complex methods such as ANP, which requires modeling interdependencies among criteria, or outranking methods such as ELECTRE and PROMETHEE, which require additional preference parameters and threshold settings, the AHP–TOPSIS framework offers greater transparency, lower computational complexity, and easier interpretation for policy-oriented decision-making. Consequently, the combination of AHP and TOPSIS provides a robust and practical approach for identifying the most suitable rural e-commerce investment priorities under multiple economic, social, and environmental criteria.

2.1 Study area, decision units, and alternatives

The empirical setting comprises a set of representative rural areas in China where e-commerce development is a policy priority and where the basic statistics needed for AHP TOPSIS scoring are available. To support meaningful comparison across different development contexts, the study area is constructed to reflect regional heterogeneity in digital infrastructure, logistics conditions, industrial structure, and income levels. In practice, this is achieved by selecting rural counties from four macro-regions: the coastal, central, western, and northeastern regions. Figure 2 presents the geographic distribution of the selected case areas and provides a visual reference for the later comparative discussion.

FIGURE 2

The selection of specific case areas follows three transparent rules. First, areas are chosen to cover different stages of rural ecommerce maturity, including early stage, growing, and relatively mature ecosystems, so that the decision framework is tested under diverse conditions. Second, areas must have sufficient data availability for the evaluation indicators, particularly for digital connectivity, logistics capacity, and socioeconomic outcomes, so that the decision matrix can be constructed with limited missing values and consistent definitions. Third, areas are chosen where public interventions are realistic and policy-relevant, for example, counties with ongoing rural revitalization programs, logistics upgrading initiatives, or e-commerce-related poverty reduction and entrepreneurship support, ensuring that the ranked alternatives correspond to implementable actions rather than purely theoretical options.

Within the selected study area, the decision units are defined as county-level rural economies, because the main public investments that enable rural ecommerce, such as last-mile logistics facilities, township distribution points, cold chain nodes, training programs, and local standards and traceability systems, are typically planned, financed, and implemented at or below the county administrative level. Using counties as decision units also aligns the evaluation with observable differences in infrastructure endowment and market access that directly influence both economic outcomes and the feasibility of interventions. For each decision unit, a consistent data template is used to populate the indicator values required by TOPSIS, while AHP is used to provide a stable weighting structure that reflects decision-makers’ preferences rather than local-scale effects.

The alternatives in the AHP-TOPSIS model are defined as mutually distinguishable public investment packages that a local government could prioritize under budget constraints. In line with the research objective, each package represents a coherent bundle of interventions rather than a single project, allowing the evaluation to capture complementarities across infrastructure, services, and capability building. Typical alternatives include a digital infrastructure upgrade package, a low-carbon logistics and consolidation package, a cold chain and warehousing package, a skills and entrepreneurship training package, and a market linkage and standardization package. These alternatives are scored against the criteria system using the county-specific indicator values, enabling the framework to rank investment priorities within each decision unit and, when needed, compare how recommended priorities vary across the case areas shown in Figure 2.

2.2 Construction of the evaluation index system

The evaluation index system is constructed to translate the research objective into a measurable hierarchy of criteria and indicators that can be weighted using the AHP and scored using the TOPSIS method. The construction procedure follows a structured synthesis of evidence and practical measurability. First, a broad pool of candidate indicators is compiled from the literature and policy practice on rural e-commerce development, digital infrastructure, logistics modernization, and rural income improvement.

In total, 12 sources are reviewed to build the initial indicator pool, including peer-reviewed journal articles and policy-oriented reports, ensuring the criteria reflect both academic consensus and implementable public investment levers. Second, the candidate pool is screened to ensure conceptual relevance and data feasibility. Conceptual relevance is ensured by retaining indicators that are directly linked to the mechanisms by which rural ecommerce affects economic outcomes, namely, reduced transaction costs, improved market access, productivity gains from digitization, and improved supply chain efficiency. Data feasibility is ensured by prioritizing indicators that can be consistently measured across the selected counties using official statistics, platform-related proxy variables, or standardized administrative records. Indicators with high missingness, inconsistent definitions, or strong overlap with other measures are removed or merged to reduce redundancy. Third, the remaining indicators are organized into a three-level hierarchical structure.

The first level is the overall goal: to prioritize public investment packages that maximize inclusive economic benefits through rural e-commerce. The second level consists of major criteria representing the key enabling and outcome dimensions. The third level consists of measurable subcriteria and indicators that can populate the decision matrix. This hierarchical structure supports the AHP weighting process by ensuring that criteria at each level are mutually distinct and collectively cover the decision problem. Fourth, indicator direction and standardization rules are defined to ensure compatibility with TOPSIS.

Each indicator is classified as a benefit-type indicator, in which larger values are preferred, or a cost-type indicator, in which smaller values are preferred. Cost type indicators are converted to a benefit-oriented form using an appropriate transformation so that all criteria follow a consistent preference direction. Units are harmonized, and normalization will be applied in the TOPSIS step to remove scale effects across indicators. Fifth, the final index system is validated through expert review.

Domain experts and local practitioners examine whether the criteria adequately represent policy controllable levers and whether the indicators can be credibly observed at the county level. Feedback is used to refine definitions, clarify measurement approaches, and confirm that the system is suitable for pairwise comparison in AHP and for quantitative scoring in TOPSIS. The resulting evaluation index system, therefore, provides the methodological bridge between the conceptual framework and the operational ranking results reported later.

2.3 Data collection and sources

Data for the AHP-TOPSIS model are assembled to ensure that each alternative is evaluated using a consistent set of indicators across the selected case areas. The data collection strategy combines secondary quantitative data with structured expert input to capture both measurable performance and contextual feasibility. Quantitative indicator values are compiled primarily from official statistical yearbooks and government open data portals at the national, provincial, municipal, and county levels, as well as from publicly available information from logistics service providers and digital infrastructure reports, where relevant. The collection focuses on variables that can be defined consistently across locations, such as connectivity coverage, logistics capacity, proxies, local economic structure, and household income-related indicators.

When indicators are not directly observable at the county level, appropriate proxy measures are adopted at the closest available administrative scale and then aligned with county decision units using consistent allocation rules, documented for transparency. Zhengzhou Urban Construction Vocational College confirmed that all questionnaires and the online protocol were completed, and that the data obtained are kept confidential to maintain the anonymity of the surveys.

After compilation, a preprocessing protocol is applied to prepare a complete TOPSIS decision matrix. First units are harmonized to avoid scale distortions, and all indicators are checked for directionality so that each variable can be treated as a benefit-oriented criterion after transformation when needed. Second, missing values are handled using a rule-based approach that prioritizes official substitute series and conservative imputation only when gaps are limited and do not affect the indicator’s interpretation.

Third, outliers and extreme values are examined because TOPSIS distances can be sensitive to large magnitudes, and a log transformation is applied for highly skewed indicators while preserving ordinal meaning. Finally, all processed values are stored in a unified dataset to enable normalization and weighting in the subsequent TOPSIS stage. Expert elicitation is used to derive criteria weights for AHP because the relative importance of economic and enabling conditions for rural e-commerce investment cannot be inferred solely from observed data. A total of sixty experts are invited to participate in a structured questionnaire survey, and each participant completes pairwise comparisons following the standard AHP judgment scale.

The recruitment strategy is designed to ensure diversity in expertise and institutional perspective so that the resulting weights reflect a balanced view of policy implementation and market realities. The expert panel includes participants from academia with research experience in regional development, the digital economy, or environmental economics; industry practitioners from e-commerce platforms, logistics, and agribusiness; technology entrepreneurs; and public sector officials involved in rural revitalization, commerce, or infrastructure planning. This diversity reduces the risk of single-stakeholder bias and enhances the external credibility of the weight system.

To ensure methodological rigor, several procedures were implemented to enhance data reliability and validity. A total of 60 experts were invited to participate in the study, representing academia, public administration, e-commerce platforms, logistics providers, agribusiness organizations, and technology enterprises. Of these, 56 questionnaires were returned, and 52 were retained after screening for completeness, reciprocity, and logical consistency. Content validity was strengthened by selecting experts with substantial professional experience in rural development, the digital economy, logistics, environmental management, and policy implementation. Construct validity was supported by developing the evaluation criteria and indicators through an extensive review of the relevant literature and consultation with domain experts. Reliability was ensured by applying the standard AHP consistency test. Individual responses and aggregated pairwise comparison matrices were examined using the Consistency Ratio (CR), and only matrices satisfying the accepted threshold of CR ≤ 0.10 were retained for analysis. Furthermore, individual judgments were aggregated using the geometric mean, which is widely recommended for group decision-making and helps reduce individual bias and improve the stability of the resulting weights.

Data are collected using a mixed-mode approach. The main questionnaire is distributed electronically with embedded guidance examples and a brief training note explaining pairwise comparisons and consistency requirements. For participants with limited access or a preference for assisted completion, the same instrument is administered in an interview-assisted format, where responses are recorded directly into the comparison matrices. Completed questionnaires are screened for completeness and internal coherence before inclusion.

Individual pairwise comparison matrices are then aggregated into a group judgment matrix using a geometric mean procedure at the element level, which is standard for combining AHP judgments from multiple experts. Consistency is checked at both the individual and aggregated levels to identify problematic submissions and ensure that the final weight vector reflects coherent preferences. Figure 3 summarizes the full elicitation process, from expert selection and invitation through questionnaire administration, judgment aggregation, and the derivation of final AHP weights, which are subsequently used in TOPSIS scoring and ranking.

FIGURE 3

2.4 Analytic hierarchy process (AHP) method

AHP is used to derive the relative importance weights of criteria and subcriteria from the pairwise comparison questionnaire completed by the expert panel. At each hierarchy level, experts compare criteria in pairs using a ratio scale, and the resulting judgments are converted into a weight vector that satisfies internal consistency requirements before being used in the TOPSIS stage. Let a hierarchy level contain criteria. Each expert provides a reciprocal pairwise comparison matrix of size , where expresses the importance of criterion relative to criterion . The matrix satisfies

Because weights are intended to represent a collective preference structure, the individual matrices are aggregated into a single group matrix using the geometric mean across the experts

This aggregation preserves reciprocity and is appropriate for ratio scale judgments. Weight derivation is then performed from the aggregated matrix. Using the geometric mean priority approximation, the unnormalized priority score for criterion is:and the normalized weight vector is obtained bywhere . This procedure is applied at each level of the hierarchy to obtain local weights. Global weights for the lowest-level indicators are computed by multiplying local weights along the unique path from the goal to each indicator. If a criterion has weight and one of its subcriteria has a conditional local weight , then the global weight of is

Consistency testing is conducted to ensure that expert judgments are sufficiently coherent for reliable weighting. First, the maximum eigenvalue of the aggregated matrix is estimated using the derived weights. Compute the vector , then estimate

The consistency index isand the consistency ratio iswhere denotes the Saaty random index for the corresponding matrix size . A judgment matrix is treated as acceptable when . When this condition is not met, the corresponding set of pairwise judgments is reviewed and corrected at the questionnaire screening stage, after which the aggregation and weight-computation steps are repeated until acceptable consistency is achieved across all hierarchy levels. The final output of this section is a validated set of global AHP weights for all indicators. These weights are then introduced directly into the TOPSIS procedure as the criterion importance parameters for constructing the weighted normalized decision matrix and producing the final ranking of alternatives.

2.5 Technique for order of preference by similarity to ideal solution (TOPSIS) method

After AHP weights are obtained and validated, TOPSIS is applied to synthesize the multi-indicator performance of each alternative and produce a final ranking. Let there be alternatives and evaluation indicators. The original decision matrix iswhere denotes the observed value of the alternative on indicator . Prior to TOPSIS, all indicators are oriented so that larger values indicate better performance. For cost-type indicators, a monotonic transformation is applied to convert them to benefit orientation, for example,ordepending on the indicator definition and interpretability. The benefit-oriented matrix is then used in subsequent steps.

2.5.1 Normalization

To remove scale effects across indicators the decision matrix is normalized using vector normalization. The normalized matrix is computed asfor and .

2.5.2 Weighted normalized decision matrix

Let denote the global weight of indicator derived from AHP and satisfying . The weighted normalized matrix is

2.5.3 Positive and negative ideal solutions

TOPSIS defines the positive ideal solution and negative ideal solution as the best and worst attainable performance across indicators within the evaluated set. They are given by

2.5.4 Separation measures

For each alternative , the Euclidean distance to the positive ideal and negative ideal solutions is computed as

2.5.5 Closeness coefficient and ranking

The relative closeness of alternative to the ideal solution is measured by the closeness coefficient where . Larger values of indicate that an alternative is closer to the positive ideal and farther from the negative ideal. The final TOPSIS ranking is obtained by sorting alternatives in descending order of .

2.6 Sensitivity analysis, robustness checks, and validation

A sensitivity analysis is conducted to evaluate how stable the TOPSIS ranking is with respect to uncertainty in the AHP-derived criterion weights. The objective is to identify whether the ordering of alternatives is driven by a small set of highly influential weights or remains consistent under plausible changes in decision-maker preferences. The baseline case uses the global AHP weight vector and produces baseline closeness coefficients for all alternatives. A one way sensitivity design is implemented by perturbing one weight at a time while holding the remaining weights proportional so that the weight sum constraint is preserved. For an indicator , the perturbed weight is defined aswhere takes values within a predetermined range such as . The remaining weights are rescaled asensuring . For each perturbed weight set the TOPSIS procedure is recomputed and the resulting changes in and in the rank order are recorded. Stability is assessed using rank correlation measures between the baseline ranking and each perturbed ranking. Using Spearman rank correlation as an example, the correlation iswhere denotes the difference between the baseline rank and the perturbed rank for alternative . High correlation values indicate that the results are robust to weight uncertainty. Robustness checks extend the sensitivity analysis by testing structured scenarios that reflect alternative policy priorities. Three policy-relevant scenarios are considered by reweighting groups of indicators within the hierarchy. A growth-oriented scenario increases the relative weights of indicators linked to income expansion and employment. An inclusion-oriented scenario increases the weights assigned to distributional access and participation-related indicators. An efficiency-oriented scenario increases weights on infrastructure utilization and logistics performance indicators. In each scenario, group weights are adjusted by a fixed multiplier and then renormalized to maintain the unit sum constraint. Scenario-specific TOPSIS rankings are then compared to the baseline to identify alternatives that remain consistently high-performing across different policy stances. Validation is performed through stakeholder interpretation of the ranking outputs. The primary validation logic is triangulation between the model results and expert expectations regarding feasibility and expected impact. After the baseline and scenario results are produced, the ranked alternatives and the main drivers of closeness coefficients are summarized and shared with a subset of participating experts and local stakeholders. Feedback is used to assess whether the ranking is considered plausible given local constraints and to interpret why certain alternatives perform better in specific case areas. This step does not change the computed rankings but strengthens the credibility of the conclusions by linking quantitative outputs to contextual reasoning and implementation realities.

3 Results and discussion

3.1 Descriptive statistics of indicators and alternatives

This subsection reports descriptive statistics for the empirical inputs used in the AHP TOPSIS evaluation to document the dataset, clarify variation across observations, and support transparent interpretation of the ranking results. The set of public investment alternatives evaluated in the decision model is defined in Table 1, while the complete indicator system, including indicator direction and measurement approach, is listed in Table 2.

For each indicator in Table 2, summary statistics are computed across the observations used to populate the TOPSIS decision matrix, including minimum, maximum, mean, median, and standard deviation. The minimum and maximum values confirm that indicator magnitudes fall within plausible bounds given their definitions. The mean and median jointly describe central tendency and indicate whether distributions are symmetric or concentrated around a typical level. The standard deviation captures dispersion and shows which indicators provide stronger discriminatory information across the evaluated observations. These statistics are reported after preprocessing so that all indicators follow a consistent preference direction as specified in Table 2.

Descriptive reporting is also used to document data completeness. For every indicator in Table 2 and every alternative in Table 1, the share of missing entries and the applied treatment rule are recorded to ensure that subsequent TOPSIS computations are based on a comparable information set. When proxy measures are used, the operational definition in Table 2 serves as the reference point for interpreting the descriptive statistics and maintaining consistency across case areas.

3.2 AHP pairwise comparison outcomes and final criteria weights

The AHP weighting results are presented in four tables that document the expert panel structure, the quality control of questionnaire responses, and the final weight system used in the TOPSIS ranking stage. The AHP hierarchy used for weighting follows the criteria system defined earlier and is organized into six main criteria, labeled C1 to C6, and 20 indicators, labeled I1 to I20. The main criteria are Digital infrastructure C1, Logistics capability C2, Human capital C3, Market linkage C4, Inclusive economic impact C5, and Environmental outcomes C6. Indicator codes and definitions are provided in Table 2, while the alternative set is provided in Table 1.

Table 3 reports the expert panel profile. The panel includes sixty participants distributed across five stakeholder groups to ensure that the weight structure reflects both implementation realities and academic and technical perspectives. The distribution indicates that no single group dominates the elicitation process, which supports balanced judgments.

Table 4 reports questionnaire processing and consistency outcomes. Of the 60 invited experts, 56 returned the questionnaire, and 52 were retained after screening for completeness and reciprocity. Group matrices were then constructed for each hierarchy level by aggregating retained responses using the geometric mean at the element level. Consistency ratios were computed for each aggregated matrix. The reported values confirm that the main criteria level and the lower-level indicator matrices satisfy the acceptance rule used in this study.

Table 5 presents the final AHP weights for the six main criteria. These weights represent the relative importance assigned by the expert panel to each criterion group in achieving the evaluation objective. Logistics capability C2 and Inclusive economic impact C5 receive the largest weights, indicating that experts view supply chain functionality and measurable economic benefits as primary determinants of effective rural ecommerce investment. Digital infrastructure C1 and Human capital C3 receive moderate weights, reflecting their roles as enabling conditions. Environmental outcomes C6 receives a smaller but non-negligible weight, showing that environmental performance is considered alongside economic objectives in the integrated assessment.

The AHP results indicate that Logistics Capability (0.220) received the highest weight, followed by Inclusive Economic Impact (0.200) and Digital Infrastructure (0.180). This finding suggests that experts consider logistics efficiency and measurable socioeconomic benefits as the most critical determinants of successful rural e-commerce development. The relatively high importance assigned to logistics reflects the central role of transportation accessibility, delivery efficiency, warehousing capacity, and cold-chain availability in connecting rural producers with broader markets.

Similarly, the substantial weight assigned to Inclusive Economic Impact highlights the importance of employment creation, income growth, and participation of low-income households in achieving sustainable rural development objectives. In contrast, Environmental Outcomes received the lowest weight (0.100), indicating that although environmental considerations remain important, experts prioritize economic and operational constraints when evaluating rural e-commerce investments. These results provide valuable insight into stakeholder preferences and demonstrate that effective rural e-commerce strategies require a balance between infrastructure development, economic inclusion, and environmental sustainability.

Figure 4 visually presents the relative importance of the six main evaluation criteria derived from the AHP analysis. Logistics Capability (0.220) received the highest weight, followed by Inclusive Economic Impact (0.200) and Digital Infrastructure (0.180). These results indicate that experts consider efficient logistics systems and measurable socioeconomic benefits as the primary drivers of successful low-carbon rural e-commerce development. Human Capital (0.160) and Market Linkage (0.140) were assigned moderate importance, reflecting their supporting role in enhancing participation and market access. Environmental Outcomes (0.100) received the lowest weight; however, this criterion remains an integral component of the evaluation framework. Overall, the results suggest that stakeholders prioritize interventions that strengthen operational efficiency and economic inclusion while maintaining environmental sustainability objectives.

FIGURE 4

The AHP results indicate that Logistics Capability (0.220) received the highest weight, followed by Inclusive Economic Impact (0.200) and Digital Infrastructure (0.180). This finding suggests that experts consider logistics efficiency and measurable socioeconomic benefits as the most critical determinants of successful rural e-commerce development. The relatively high importance assigned to logistics reflects the central role of transportation accessibility, delivery efficiency, warehousing capacity, and cold-chain availability in connecting rural producers with broader markets. Similar conclusions have been reported by Du et al. (2023), who found that logistics efficiency and digitalization are important mechanisms through which rural e-commerce contributes to economic and environmental performance. Likewise, He et al. (2025) emphasized the importance of logistics optimization, resource sharing, and industrial upgrading in supporting low-carbon development. The substantial weight assigned to Inclusive Economic Impact is also consistent with studies by Zhang et al. (2024), which identified income growth, employment generation, and improvements in rural welfare as the primary objectives of rural e-commerce development. In contrast, Environmental Outcomes received the lowest weight (0.100), indicating that although environmental considerations remain important, experts prioritize economic and operational constraints when evaluating rural e-commerce investments. These findings are consistent with previous studies showing that policymakers often place greater emphasis on immediate socioeconomic benefits while pursuing long-term sustainability objectives.

Table 6 provides the global indicator weights for I1 to I20, which are used directly in the TOPSIS weighted normalized decision matrix. Each indicator weight is the product of its local weight within its criterion group and the corresponding main criterion weight in Table 5, thereby ensuring coherence across hierarchy levels. The distribution in Table 6 shows that indicators under Inclusive economic impact and Logistics capability account for a substantial share of total weight, while Digital infrastructure, Human capital, Market linkage, and Environmental outcomes remain influential through multiple indicators with moderate weights. These global weights are then used to compute the TOPSIS closeness coefficients and the final ranking.

3.3 TOPSIS closeness coefficients and baseline ranking of alternatives

This subsection reports the baseline TOPSIS results produced using the global AHP indicator weights in Table 6 and the TOPSIS procedure defined in Section 2.5. After orientation, alignment, and normalization, the weighted normalized decision matrix was used to derive the positive and negative ideal solutions. For each alternative, the Euclidean separation from the positive ideal and from the negative ideal were computed across the full indicator set to . The closeness coefficient was then calculated as . A larger indicates that an alternative is closer to the positive ideal and farther from the negative ideal, leading to a higher baseline rank.

Table 7 reports the separation measures and closeness coefficients for each alternative. The baseline results show that has the smallest distance to the positive ideal and the largest distance to the negative ideal , producing the highest closeness coefficient . This outcome indicates that under the baseline weighting structure, the performance profile of aligns most closely with the best attainable values across the indicator set while remaining farthest from the worst-case profile. The second-ranked alternative is with , , and . Although its distance to the positive ideal is slightly larger than , the relatively high suggests that avoids weak performance across multiple weighted indicators, thereby improving its overall closeness.

The middle-ranking alternatives and have moderate closeness coefficients of and , respectively. Their distances to the positive ideal remain higher than those of and , with and , indicating that their indicator profiles are less aligned with the best observed weighted performance across the full set of criteria. At the same time, their distances to the negative ideal are not as large as those of the top ranked alternatives, with and , implying the presence of comparatively weaker dimensions that pull them closer to the negative ideal point.

The lowest ranked alternative is with and , which yields the smallest closeness coefficient . This result is driven by two features visible in Table 7. First, has the largest separation from the positive ideal among all alternatives, suggesting weaker alignment with the best observed performance on several indicators that carry nontrivial weights in Table 6. Second, has the smallest separation from the negative ideal, indicating that its performance profile is comparatively closer to the worst-case values on a subset of indicators. To clarify the magnitude of differences between alternatives, Table 8 restates the ranking and highlights gaps in the closeness coefficients. The difference between the first and second-ranked alternatives is 0.630–0.593 = 0.037, while the gap between the second and third-ranked alternatives is 0.593–0.556 = 0.037. The smaller gap between the third and fourth is 0.556–0.534 = 0.022, indicating that and are more similar in overall performance under the baseline weights. The gap between fourth and fifth is 0.534–0.489 = 0.045, showing a clearer separation between and the remaining alternatives. Overall, the baseline TOPSIS results identify a top tier consisting of and , a middle tier consisting of and , and a lower tier represented by . This baseline ranking serves as the reference case for the sensitivity and scenario analyses in later subsections, which evaluate whether the ordering remains stable when the AHP weight structure is perturbed or when alternative policy-priority scenarios are imposed.

The baseline ordering implied by Table 7 is ranked first, ranked second, ranked third, ranked fourth, and ranked fifth. For clarity, Table 8 restates the ranking in a compact form, along with the corresponding closeness coefficients.

The ranking results indicate that Skills and Entrepreneurship Support (A4) achieved the highest overall performance, suggesting that investments in human capital development, digital literacy, business incubation, and entrepreneurial capabilities generate benefits across multiple dimensions simultaneously. The strong performance of A4 demonstrates that infrastructure investments alone may not be sufficient unless rural populations possess the skills required to effectively utilize digital platforms and market opportunities. Low-Carbon Last-Mile Logistics (A2) ranked second, reflecting the importance of efficient distribution systems in improving market access while reducing environmental impacts.

By contrast, Warehousing and Cold Chain Infrastructure (A3) obtained the lowest ranking, indicating that although storage and preservation facilities remain important, their benefits are comparatively limited when broader digital, human capital, and logistics constraints remain unresolved. The ranking, therefore, provides a practical prioritization strategy for policymakers seeking to maximize the effectiveness of limited rural development resources.

Figure 5 presents the TOPSIS closeness coefficients for the five investment alternatives. The results show that A4 achieved the highest closeness coefficient (0.630), indicating the strongest overall performance across the weighted economic, social, logistical, and environmental criteria. A2 ranked second with a coefficient of 0.593, demonstrating the importance of low-carbon logistics investments in improving both operational efficiency and sustainability outcomes. A5 (0.556) and A1 (0.534) formed a middle-performance group with relatively similar scores, while A3 obtained the lowest coefficient (0.489), indicating weaker overall alignment with the ideal solution. The results suggest that investments focused on skills development, entrepreneurship support, and low-carbon logistics offer the most balanced and effective pathways to promote sustainable rural e-commerce development.

FIGURE 5

3.4 Regional comparison of rankings across case areas and key criteria driving results

Here, we compare how the baseline TOPSIS ranking varies across the selected case areas and identify the criteria that most strongly drive those regional differences. The comparison is implemented by computing TOPSIS closeness coefficients and ranks separately for each case area, using the same global AHP weights in Table 6, so that observed variation reflects differences in indicator performance rather than changes in the preference structure. Results are then grouped by macro-region to highlight systematic patterns across the coastal, central, western, and northeastern cases. Table 9 reports the within-region rankings, while Table 10 summarizes the dominant criteria that explain why top-ranked alternatives differ across regions.

Across coastal case areas, the highest-ranked alternatives are typically those that strengthen market linkage and low-carbon logistics. This pattern is consistent with coastal contexts where basic connectivity is often less binding and marginal benefits come from improving supply chain efficiency and access to larger markets. In these areas, alternatives that reduce delivery time, increase logistics accessibility, and improve platform access tend to move closer to the positive ideal, particularly when the corresponding indicators carry moderate to high weights through Logistics capability and Inclusive economic impact in Tables 5, 6. By contrast, in Western case areas, rankings more frequently place digital infrastructure upgrades, skills, and entrepreneurship support in the top tier. This reflects weaker baseline connectivity and lower digital readiness, where improvements in broadband coverage affordability and training participation produce larger relative gains across multiple indicators and thus yield larger reductions in distance to the positive ideal solution. Central case areas often exhibit mixed profiles in which logistics and market linkages compete with human capital interventions, depending on existing distribution networks and labor mobility conditions. Northeastern cases frequently show stronger sensitivity to market linkage and product standardization because revitalization objectives and industrial structure can make access to stable demand channels and compliance-related capacities more influential for income growth outcomes.

To determine which criteria drive these regional outcomes, the contribution of each main criterion to the TOPSIS separation measures is examined by decomposing the weighted squared distance terms. For alternative , the criterion level contribution to separation from the positive ideal can be expressed as the share of the total squared distance attributable to indicators within each criterion group. In practice, this is computed by summing the squared deviations over indicators belonging to the same criterion and dividing by the total . A similar decomposition is computed for the negative ideal distances. The resulting shares reveal whether an alternative ranks highly because it performs strongly on Logistics capability and Inclusive economic impact indicators, or because it achieves balanced improvements across enabling dimensions such as Digital infrastructure, Human capital, and Market linkage.

Table 10 reports the dominant criteria drivers by region based on the average contribution shares among the top-ranked alternatives in each macro region. The results indicate that Logistics capability and Inclusive economic impact are the most frequent primary drivers in coastal cases, while Digital infrastructure and Human capital are more influential in western cases. Market linkage emerges as a consistent secondary driver across all regions, particularly in case areas where product standardization and traceability adoption remain low. Environmental outcomes contribute meaningfully in cases where delivery emissions intensity and packaging waste pressure are relatively high because those indicators can increase distance to the positive ideal even when economic indicators are strong. Consequently, alternatives emphasizing low carbon logistics can improve rankings in regions with high baseline transport intensity.

3.5 Sensitivity analysis scenario-based ranking changes and validation summary

This section evaluates the stability of the baseline ranking under changes in the AHP weight structure and summarizes how experts and stakeholders interpret the robustness of the results. The analysis is organized into three parts. First, one-way weight perturbations are applied to assess the TOPSIS ranking’s sensitivity to uncertainty in individual criterion importance. Second, structured policy scenarios are constructed to represent alternative priority settings, and the resulting shifts in ranking are examined. Third, a validation summary is reported based on stakeholders’ interpretations of the outputs and the consistency of the results with implementation realities.

Sensitivity analysis is conducted by perturbing one criterion weight at a time while maintaining the unit sum constraint through proportional rescaling of the remaining weights. The perturbation range is set to ±30% of the baseline weight for the selected criterion. For each perturbation run, the TOPSIS procedure is recomputed, and rank changes relative to the baseline ordering are recorded. Table 11 summarizes the rank stability outcomes using the number of rank reversals and the Spearman rank correlation between the perturbed ranking and the baseline ranking. Overall, the results indicate that the top-ranked alternative remains stable across most perturbations. Ranking changes primarily occur within the middle tier, where alternatives have closer baseline closeness coefficients, as shown previously. The most influential perturbations are those applied to Logistics capability and Inclusive economic impact because these criteria carry larger baseline weights and include several indicators with strong cross-alternative dispersion. When the weight on Logistics capability is increased, the ranking of improves relative to in some runs, reflecting the stronger alignment of with logistics efficiency-related indicators. When the weight on Human capital is increased, tends to remain strong while can move upward in Western-like profiles where training and literacy proxies are comparatively weak at baseline.

Scenario analysis complements the one-way perturbations by constructing policy relevant reweighting schemes that represent different government objectives. A growth-oriented scenario increases the combined weight assigned to Inclusive economic impact and Market linkage, an inclusion-oriented scenario increases the combined weight assigned to Human capital and participation-related indicators within Inclusive economic impact, and an environment-oriented scenario increases the weight assigned to Environmental outcomes together with low-carbon logistics-related indicators. For each scenario, weights are renormalized, and the TOPSIS ranking is recomputed. Table 12 reports the resulting rank order for each scenario and identifies the alternatives that remain in the top two positions across scenarios. The results show that remains top-ranked in the baseline and inclusion-oriented scenarios, while cap A. 2strengthens under the environment-oriented scenario because low-carbon logistics improvements simultaneously affect delivery emissions intensity and logistics efficiency. Under the growth-oriented scenario and typically improve because enhanced market linkage and logistics performance translate more directly into income and employment proxies. Robust alternatives are defined as those that remain in the top two across at least three of four settings, including the baseline. Under this definition, and are identified as robust investment priorities.

Validation is implemented through structured interpretation of the results with experts and local stakeholders. A subset of respondents from the expert panel reviewed the baseline and scenario rankings together with the main drivers derived from criterion contribution analysis. The validation feedback indicates that the identified top-tier alternatives are considered feasible and consistent with the observed bottlenecks in rural e-commerce systems. Stakeholders emphasized that improvements in skills and entrepreneurship support can unlock the benefits of infrastructure and market access, which aligns with the persistent high ranking of . They also noted that low-carbon logistics investments are increasingly prioritized due to both cost pressures and environmental compliance expectations, supporting the strong performance of in the environment-oriented scenario. Where disagreements occurred, they primarily concerned the relative ordering of middle-tier alternatives rather than the identification of the highest-priority options, consistent with the previously observed small closeness coefficient gaps. Overall, the combined sensitivity analysis and validation evidence support the credibility of the baseline conclusions and indicate that the main policy recommendation is robust to plausible changes in weighting priorities.

3.6 Discussions

Across the robustness assessments, the baseline ordering remains largely stable, and the main conclusion that the top tier consists of A4 and A2 is sustained under both univariate weight perturbations and structured policy scenarios. Under one-way perturbations of each main criterion weight by plus or minus 30 percent, with proportional rescaling of the remaining weights, the number of rank reversals is limited and concentrated in the middle tier rather than the top tier, as summarized in Table 11.

The findings are broadly consistent with recent studies emphasizing the importance of human capital, digital readiness, and logistics efficiency in rural e-commerce development. For example, Ji et al. (2023); Du et al. (2023) reported that improvements in logistics systems and digitalization contribute significantly to both economic development and carbon emission reduction. Similarly, He et al. (2025) found that e-commerce development promotes low-carbon growth through productivity enhancement, industrial upgrading, and resource-sharing mechanisms.

The strong performance of A2 in the present study supports these conclusions by demonstrating the importance of logistics capability as a key driver of sustainable rural development. Likewise, A4’s top ranking highlights the critical role of skills development and entrepreneurship support, extending previous findings that digital infrastructure alone cannot guarantee successful participation in e-commerce ecosystems.

The study also helps address several gaps identified in the literature. Previous research has primarily focused on estimating the economic or environmental impacts of rural e-commerce programs, whereas limited attention has been given to the practical prioritization of alternative public investment strategies. Furthermore, existing studies often examine economic and environmental outcomes separately, making it difficult for policymakers to evaluate trade-offs across multiple objectives. The integrated AHP–TOPSIS framework developed in this study bridges these gaps by simultaneously incorporating economic, social, logistical, and environmental criteria within a single decision-support model. The inclusion of sensitivity and scenario analyses further extends existing research by identifying investment priorities that remain robust under different policy preferences, thereby providing more actionable guidance for rural development planning.

Changes to Logistics capability and Inclusive economic impact generate the largest instability, with two rank reversals and a Spearman rank correlation as low as 0.80, indicating that these criteria are the primary levers through which preference uncertainty can alter ordering. This sensitivity is consistent with their larger baseline weights in Table 5 and with the fact that their associated indicators represent operational and outcome dimensions that vary markedly across alternatives. By contrast, perturbations to Digital infrastructure, Human capital, Market linkage, and Environmental outcomes lead to at most one rank reversal and maintain rank correlations within 0.90–1.00, suggesting that moderate shifts in these weights do not substantially change the overall preference structure.

Scenario analysis further clarifies how rankings respond to policy emphasis, as shown in Table 12. In the growth-oriented scenario, A2 rises to rank 1, A5 to rank 2, and A4 to rank 3, indicating that when income and market expansion proxies are prioritized, the relative advantage of logistics and market linkage packages strengthens. In the inclusion-oriented scenario, A4 remains rank 1, and A1 rises to rank 2, reflecting that a heavier emphasis on participation and capability building rewards alternatives that directly improve human capital and access, enabling conditions. In the environment-oriented scenario, A2 ranks first, and A4 remains second, consistent with the dual effect of low-carbon logistics on delivery emissions intensity and operational efficiency. Numerically, the baseline closeness coefficients reported earlier show a top-tier separation with A4 at 0.630 and A2 at 0.593, followed by a middle-tier A5 at 0.556, and A1 at 0.534, and a lower-tier A3 at 0.489, and these gaps help explain why rank changes occur mainly between adjacent middle alternatives rather than displacing the leading options.

Validation feedback from experts and stakeholders supports these patterns by emphasizing that skills and entrepreneurship support is a binding complement that unlocks the returns from connectivity logistics and platform access, aligning with the persistent strength of A4, while low carbon logistics is increasingly viewed as both a cost and compliance priority, aligning with the scenario-driven improvement of A2. Future work should extend the analysis in three directions by incorporating dynamic effects through multi period data to capture how investments shift indicators over time integrating uncertainty explicitly through fuzzy AHP or probabilistic TOPSIS to represent expert disagreement and measurement error and expanding the alternative set to include combined packages and budget constrained portfolio selection so that the model can recommend not only the best single priority but an implementable mix of interventions under realistic fiscal limits.

Although the empirical analysis focuses on China, the findings have broader relevance to developing economies promoting digital transformation in rural areas. Many countries face similar challenges, including inadequate logistics systems, uneven digital access, limited entrepreneurial capabilities, and growing environmental concerns linked to e-commerce expansion. The integrated AHP–TOPSIS framework developed in this study provides a transferable decision-support approach to assist policymakers in evaluating and prioritizing rural e-commerce investments across multiple economic, social, and environmental objectives. Therefore, the study’s contribution extends beyond the Chinese context and offers methodological insights applicable to other emerging and developing economies.

3.7 Theoretical implications

The findings of this study are broadly consistent with previous research emphasizing the importance of logistics capability, digital infrastructure, and human capital in rural e-commerce development. Du et al. (2023) demonstrated that improvements in logistics efficiency and digitalization contribute significantly to both economic development and carbon emission reduction. Similarly, He et al. (2025) reported that the development of e-commerce promotes low-carbon growth through industrial upgrading, resource sharing, and innovation mechanisms. The relatively high weights assigned to Logistics Capability and Inclusive Economic Impact in the present study support these findings and confirm that operational efficiency and socioeconomic benefits remain central considerations in rural e-commerce investment decisions.

The results also extend existing knowledge by highlighting the importance of skills development and support for entrepreneurship. The highest ranking achieved by A4 indicates that human capital investments can generate broader and more balanced benefits than infrastructure-only interventions. This finding complements earlier studies by Zhang et al. (2024), which emphasized the role of rural e-commerce in income growth and employment creation but offered limited guidance on prioritizing alternative investment strategies. The present study suggests that the effectiveness of digital infrastructure and logistics investments depends substantially on the capacity of rural populations to utilize these opportunities productively.

The study further contributes to the literature by addressing several important research gaps. Previous studies have primarily focused on estimating the economic or environmental impacts of rural e-commerce programs, while limited attention has been paid to prioritizing competing public investment alternatives. Moreover, economic and environmental outcomes have often been examined separately. The integrated AHP–TOPSIS framework developed in this study bridges these gaps by simultaneously incorporating economic, social, logistical, and environmental criteria within a unified decision-support system. The inclusion of sensitivity and scenario analyses further strengthens the framework’s practical relevance by identifying investment priorities that remain robust across different policy objectives and stakeholder preferences.

From a policy perspective, the findings provide evidence-based guidance for allocating limited rural development resources. The results indicate that investments in skills and entrepreneurship support, and in low-carbon logistics systems, offer the greatest potential to achieve balanced economic and environmental outcomes. Therefore, the study contributes not only to the academic literature on rural digital transformation and sustainable development but also to practical decision-making processes in China and other developing economies facing similar rural development challenges.

4 Conclusion

An integrated AHP-TOPSIS decision framework was applied to prioritize public investment options for low-carbon rural e-commerce development across multiple criteria and indicators. Using aggregated pairwise comparison judgments from sixty experts, the final weight system was combined with the TOPSIS ideal solution approach to derive baseline rankings and to test stability under preference uncertainty and alternative policy priorities.

In the baseline case, the top tier consisted of A4 and A2. A4 achieved the highest closeness coefficient with a relatively small distance to the positive ideal and the largest distance to the negative ideal . A2 ranked second with . The middle tier included A5 with and A1 with indicating similar overall performance under the baseline weights, with a small gap of 0.022 between them. A3 ranked last with showing the weakest alignment with the positive ideal and the closest proximity to the negative ideal among the alternatives.

Robustness checks reinforce the reliability of the main conclusion. Under one-way perturbations of criterion weights by ±30%, the overall ranking remained stable at the top with limited rank reversals concentrated in the middle tier. The greatest sensitivity occurred when the Logistics capability and Inclusive economic impact weights were perturbed with up to two rank reversals, and Spearman’s rank correlation was as low as 0.80, while other criteria perturbations generally maintained correlations between 0.90 and 1.00. Scenario analysis produced intuitive ranking shifts. A2 moved to rank 1 in the growth-oriented and environment-oriented settings, while A4 remained rank 1 in the inclusion-oriented setting and stayed within the top two across all tested scenarios. Overall, the results support a practical investment implication. Interventions that build human capital and entrepreneurship capacity and those that improve last-mile logistics with low-carbon features yield the most consistently favorable multi-criteria performance. Future work should extend the framework by incorporating multiyear data to capture dynamic impacts using uncertainty-aware weighting approaches, such as fuzzy judgments, and expanding from single option ranking to portfolio selection under explicit budget constraints, so that optimal combinations of interventions can be recommended.

References

• 1

  BachofnerM.LemardeléC.EstradaM.PagèsL. (2022). City logistics: challenges and opportunities for technology providers. J. Urban Mobility2, 100020. 10.1016/j.urbmob.2022.100020

  • CrossRef
  • Google Scholar
  • View reference in article

• 2

  ChenC.PanJ. (2019). The effect of the health poverty alleviation project on financial risk protection for rural residents: evidence from Chishui City, China. Int. J. Equity Health18 (1), 79. 10.1186/s12939-019-0982-6

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 3

  ChuF.ZhangW.JiangY. (2021). How does policy perception affect green entrepreneurship behavior? An empirical analysis from China. Discrete Dyn. Nat. Soc.2021 (1), 7973046–7973049. 10.1155/2021/7973046

  • CrossRef
  • Google Scholar
  • View reference in article

• 4

  DingS.CaoS.SunY.ZhaoS.XuB.LiuD. (2025). Health burden and associated economic losses attributable to atmospheric black carbon in Hangzhou, East China: characteristics, source apportionment and policy implications. Atmos. Environ.363, 121615. 10.1016/j.atmosenv.2025.121615

  • CrossRef
  • Google Scholar
  • View reference in article

• 5

  DuL.WangX.ZhangS. (2023). Can rural e-commerce contribute to carbon reduction? A quasi-natural experiment based on China’s e-commerce demonstration counties. Environ. Sci. Pollut. Res.30 (47), 104336–104349. 10.1007/s11356-023-29792-z

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 6

  HeX.MaD.TangL. (2025). How E-Commerce drives low-carbon development: an empirical analysis from China. Sustainability17 (19), 8818. 10.3390/su17198818

  • CrossRef
  • Google Scholar
  • View reference in article

• 7

  HuJ.YinG. (2026). How rural E-Commerce shapes agricultural carbon emissions: evidence from a quasi-natural experiment in China. Sustainability18 (11), 5251. 10.3390/su18115251

  • CrossRef
  • Google Scholar
  • View reference in article

• 8

  HuF.YangH.ZhouX.ZhaoS.QiuL.WeiS.et al (2025). Spatial dynamics of lithium battery recycling enterprises in China: implications for smart waste management and public health. Front. Public Health13, 1729413. 10.3389/fpubh.2025.1729413

  • CrossRef
  • Google Scholar
  • View reference in article

• 9

  HuF.YangH.WeiS.ZhouH.ChenY.HuH. (2026). Urban green technology transfer networks and green finance development: evidence from the Yangtze River Delta, China. Pacific-Basin Finance J.96, 103055. 10.1016/j.pacfin.2026.103055

  • CrossRef
  • Google Scholar
  • View reference in article

• 10

  JiK.HouQ.YuY.PanD. (2023). Rural E-Commerce and agricultural carbon emission reduction: a quasi-natural experiment from China’s rural E-Commerce demonstration County program based on 355 cities in ten years. Agriculture14 (1), 75. 10.3390/agriculture14010075

  • CrossRef
  • Google Scholar
  • View reference in article

• 11

  JiangS.YuanX. (2026). Does the e-CNY pilot boost consumption upgrading? Evidence from China’s city-level quasi-experiment. Appl. Econ. Lett.1 (1), 1–5. 10.1080/13504851.2026.2634858

  • CrossRef
  • Google Scholar
  • View reference in article

• 12

  JiangT.SunQ.YanQ. (2026). The impact of normative misperception on low-carbon travel: a chain mediation model. Soc. Behav. Personal.54 (3), 1–17. 10.2224/sbp.14977

  • CrossRef
  • Google Scholar
  • View reference in article

• 13

  KangP.SongG.XuM.MillerT. R.WangH.ZhangH.et al (2021). Low-carbon pathways for the booming express delivery sector in China. Nat. Commun.12 (1), 450. 10.1038/s41467-020-20738-4

  • CrossRef
  • Google Scholar
  • View reference in article

• 14

  KumarN.TyagiM.GargR. K.SachdevaA.PanchalD. (2020). “A framework development and assessment for cold supply chain performance system: a case of vaccines,” in Operations Management and Systems Engineering: Select Proceedings of CPIE 2019 (Singapore, Singapore: Springer), 339–353.

  • Google Scholar
  • View reference in article

• 15

  KumarN.TyagiM.SachdevaA. (2022). Depiction of possible solutions to improve the cold supply chain performance system. J. Adv. Manag. Res.19 (1), 106–138. 10.1108/jamr-10-2020-0285

  • CrossRef
  • Google Scholar
  • View reference in article

• 16

  KumarN.TyagiM.SachdevaA. (2023). Technological assessment of global warming inference for cold supply chain performance. Int. J. Logist. Syst. Manag.44 (1), 17–45. 10.1504/ijlsm.2020.10053022

  • CrossRef
  • Google Scholar
  • View reference in article

• 17

  KumarN.TyagiM.SachdevaA. (2025). Interval-valued Pythagorean fuzzy-based decision support framework for ensuring the food quality through enhanced cold supply chain performance system. Qual. Manag. J.32 (4), 208–238. 10.1080/10686967.2025.2564638

  • CrossRef
  • Google Scholar
  • View reference in article

• 18

  LiG.QinJ. (2022). Income effect of rural E-commerce: empirical evidence from Taobao villages in China. J. Rural Stud.96, 129–140. 10.1016/j.jrurstud.2022.10.019

  • CrossRef
  • Google Scholar
  • View reference in article

• 19

  LiuY.ZhouM. (2023). Can rural e-commerce narrow the urban–rural income gap? Evidence from coverage of Taobao villages in China. China Agric. Econ. Rev.15 (3), 580–603. 10.1108/caer-09-2022-0221

  • CrossRef
  • Google Scholar
  • View reference in article

• 20

  LuF.YoonS. J. (2026). Cultural–tourism integration and people’s livelihood and well-being in China’s Yellow River Basin: dynamic Panel evidence and spatial spillovers (2011–2023). Sustainability18 (2), 1006. 10.3390/su18021006

  • CrossRef
  • Google Scholar
  • View reference in article

• 21

  LuoX.NiuC. (2019). “E-commerce participation and household income growth in Taobao villages,” Report No.: 8811. Washington, DC, United States: World Bank Policy Research Working Paper. 10.1596/1813-9450-8811

  • CrossRef
  • Google Scholar
  • View reference in article

• 22

  LuoZ.-H.ZhuC.-Z. (2025). Study on the spatial pattern of the carbon footprint of China’s e-Commerce express packaging considering embodied carbon transfer. Sustainability17 (11), 5102. 10.3390/su17115102

  • CrossRef
  • Google Scholar
  • View reference in article

• 23

  LuoJ.WangY.LiG. (2023). The innovation effect of administrative hierarchy on intercity connection: the machine learning of twin cities. J. Innov. Knowl.8 (1), 100293. 10.1016/j.jik.2022.100293

  • CrossRef
  • Google Scholar
  • View reference in article

• 24

  LvJ.GuoX.JiangH. (2025). Rural e-Commerce and income inequality: evidence from China. Sustainability17 (10), 4720. 10.3390/su17104720

  • CrossRef
  • Google Scholar
  • View reference in article

• 25

  NwabekeeU. S.OgeawuchiJ. C.AbayomiA. A.AgboolaO. A.GeorgeO. O. (2023). A conceptual framework for data-informed gig economy infrastructure development in last-mile delivery systems. J. Front. Multidiscip. Res.4 (2), 82–97. 10.54660/.jfmr.2023.4.2.82-97

  • CrossRef
  • Google Scholar
  • View reference in article

• 26

  ParkJ.SarkisJ.WuZ. (2010). Creating integrated business and environmental value within the context of China’s circular economy and ecological modernization. J. Clean. Prod.18 (15), 1494–1501. 10.1016/j.jclepro.2010.06.001

  • CrossRef
  • Google Scholar
  • View reference in article

• 27

  QiaoC. (2025). Coupling digital inclusive finance and rural E-Commerce: a systems perspective on China’s urban–rural income gap. Systems13 (10), 911. 10.3390/systems13100911

  • CrossRef
  • Google Scholar
  • View reference in article

• 28

  QiuyunL.XieM.TangY. (2025). Do the E-commerce-induced business clusters increase the income of local rural households?Econ. Analysis Policy86, 1510–1526. 10.1016/j.eap.2025.04.024

  • CrossRef
  • Google Scholar
  • View reference in article

• 29

  SinghR.RaniS.SinghA.LealW.KumarV. (2026). “Smart technologies for a sustainable future: enabling reuse and recycling in the digital circular economy,” in Data-Driven Sustainability: Harnessing Technology for A Greener Future and Environmental Resilience. Editors SinghR.Leal FilhoW.KumarV. (Cham: Springer Nature Switzerland), 265–278. 10.1007/978-3-032-13412-7_16

  • CrossRef
  • Google Scholar

• 30

  SuY.DuanH.WangZ.SongG.KangP.ChenD. (2020). Characterizing the environmental impact of packaging materials for express delivery via life cycle assessment. J. Clean. Prod.274, 122961. 10.1016/j.jclepro.2020.122961

  • CrossRef
  • Google Scholar
  • View reference in article

• 31

  SunX.Abd KadirH. B. (2026). The influence of corporate social responsibility on corporate sustainable development: corporate environmental strategy in agricultural firms in Guangxi. J. Chin. Hum. Resour. Manag.17 (1), 176–192. 10.47297/wspchrmWSP2040-800510.20261701

  • CrossRef
  • Google Scholar
  • View reference in article

• 32

  TangG.ChenY.JiangY.PailléP.JiaJ. (2018). Green human resource management practices: scale development and validity. Asia Pac. J. Hum. Resour.56 (1), 31–55. 10.1111/1744-7941.12147

  • CrossRef
  • Google Scholar
  • View reference in article

• 33

  TangJ.ZhangN.WangL. (2025). Rural E-commerce policy and environmental pollution: evidence from China's poverty-stricken counties. J. Environ. Manag.393, 127061. 10.1016/j.jenvman.2025.127061

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 34

  YuW.WuY.TanX.GuoX. (2024). The nexus between electronic commerce and environmental pollution: the roles of energy efficiency and resource consumption. Bus. Strategy Environ.33 (3), 2287–2300. 10.1002/bse.3598

  • CrossRef
  • Google Scholar
  • View reference in article

• 35

  ZhangY.LongH.MaL.TuS.LiY.GeD. (2022). Analysis of rural economic restructuring driven by e-commerce based on the space of flows: the case of Xiaying village in central China. J. Rural Stud.93, 196–209. 10.1016/j.jrurstud.2018.12.001

  • CrossRef
  • Google Scholar

• 36

  ZhangL.MalekJ. A.HussainM. Y.LiaoJ. (2024). Identifying the characteristics of rural regions in China where e-commerce can effectively drive rural revitalisation. Int. J. Serv. Stand.14 (3), 240–267. 10.1504/ijss.2024.143431

  • CrossRef
  • Google Scholar

• 37

  ZhangY.WuX.CaiR. (2025). The effects of rural E-Commerce on income inequality: evidence from the national rural E-Commerce comprehensive demonstration policy in China. Econ. Analysis Policy87, 221–234. 10.1016/j.eap.2025.05.052

  • CrossRef
  • Google Scholar

• 38

  ZhangM.PanichakarnB.ShenS. (2025). Impact of green human resource management and green supply chain management practices on sustainable organizational Performance-A case of China-Laos railway. J. Chin. Hum. Resour. Manag.16 (2), 3–21. 10.47297/wspchrmWSP2040-800501.20251602

  • CrossRef
  • Google Scholar

• 39

  ZhouK.SongY.XianX.LiuY.FangW.LiuD.et al (2024). A comprehensive comparison of the life-cycle environmental impacts of traditional and returnable express delivery packaging in China. J. Clean. Prod.434, 140017. 10.1016/j.jclepro.2023.140017

  • CrossRef
  • Google Scholar